Intel VTune(TM) Performance Analyzer 3.0 Beta Test Archive

[MARIA_M_Intel]

This thread is an archive of the postings that were posted during the beta test program for this software. The product is now currently available for sale or evaluation at:

http://www.intel.com/software/products/vtune/vlin

Message Edited by Judy on 02-23-2005 03:37 PM

Message Edited by Judy on 02-23-2005 03:37 PM

[TimP]

problem was caused by confusion with multiple copies of kernel. Red Hat compounds the confusion by supplying 2 kernels for some EM64T platforms. The plain EM64T kernel which is not identified as smp actually is SMP and the Vtune 3.0 driver build worked with it.
I'm happy to run remote collector from Windows 2000, given that Red Hat doesn't support the video card or mouse on my EM64T.

Message Edited by tim18 on 08-25-2004 08:36 AM

[TimP]

Not a new problem, but I thought that rdc should be able to terminate the application when it finishes sampling. Now there are so many more types of samples to collect, it takes over an hour to complete multiple sampling runs, and it's difficult to arrange the application to terminate on its own shortly after sampling ends and Vtune starts a new copy of the application.

[David_A_Intel1]

Is the sampling collector configured to terminate the application upon completion of the activity?

[TimP]

Dave,

Yes, I leave the terminate on completion box checked, but I don't know what effect it should have; I've never noticed it to do anything. I have to watch the remote application and kill each instance when the next one starts. It has always been this way, so I guess I shouldn't mention it.

I'm still trying to figure out whether I should be running remote collector, if so, should I use Windows2000 or XP. If not, should it be done with vtl and vnc, neither of which I'm familiar with. Hard to believe after spending so much time with Vtune that it still doesn't become routine.

[jeffrey-gallagher]

Hey Tim,

This is the VTune for Linux beta forum, version 3.0. Is your question about this software, or, VTune 7.2 beta?

(On the Linux side, you can use the ActivityController to stop sampling activities.)

jdg

[TimP]

This message comes up in Vtune 7.1A when the 3.0 rdc finishes data collection. I removed the Vtune installation from Windows2000, installed it on XP, went through the 2 hours of data collection, and the same thing happened on XP. The analysis progress bar comes up under Windows 2000, but it sticks at 0%. I can't even find anything about this most frequently occurring error message in the help menu.

A hint from a colleague is that the Windows Vtune GUI exhibits this failure any time there is a message "selected event did not occur," so I am attempting a repeat. The Windows GUI deletes the data collection directories when it fails, so the evidence is gone.

Message Edited by tim18 on 08-25-2004 02:43 PM

Another error I made is that I checked the "start with data collection suspended" box, which no longer is the right thing to do when all you want is to specify a start delay.

Message Edited by tim18 on 08-25-2004 03:02 PM

[David_A_Intel1]

Tim:

If you are having a technical problem, please submit an issue at Intel Premier Support. We provide help with technical problems at the Premier web site and would be glad to try to help you.

[wayne-wooten]

I would REALLY love to have vTune3
work under FedoraCore2. I have downloaded
the course to the 2.6.6 kernel, installed
the patches from the 2.6.x driver kit
(they are badly out of date by the way,
and I had to modify the patch somewhat).

I then got the kernel to work, but I
cannot get the driver to build. First
off it is looking for ksyms.c in
/usr/src/linux-2.6.x/kernel and it is no
longer there it has been moved.

After I pointed it at the correct location,
it builds a vtune_drv.o, but when I try
to insmod that, it complains:

-1 Invalid module format

and dmesg reports

No module found in object.

So I am stuck, very close, but stuck.
I assume the module format has changed
since this module kit was released almost
a year ago. Can someone update this kit
or give me a hint to where I should look
to make the changes myself?

Thanks!
--Wayne Wooten
Pixar Animation Studios

[jeffrey-gallagher]

Hey Wayne,

We're checking this out on the engineering side now: stay tuned.

cheers

jdg

[pwlivermore]

Can someone clarify the license agreement that accompanies the beta program - in particular, can it be used for university research by salaried staff, normally not classified as "non commerical" by Intel?

[David_A_Intel1]

What do you get when you do 'vtl view -modules'? Does it show the 'vtundemo' module?

If you believe there is a problem, please submit an issue at the Intel Premier Support web site, including a pack and go file of the results (compressed).

Thanks.

[David_A_Intel1]

Did you see something in the license to make you think it could not be used in a commercial manner? If you are concerned about something specific, we can look into it. Otherwise, we do not want to give general statements that might be misinterpreted.

[David_A_Intel1]

Here is what I found out: the beta EULA does not prevent the use of the software for commercial purposes, however, there are restrictions regarding how the software may be used:

"UNAUTHORIZED USE: THE MATERIALS ARE NOT DESIGNED, INTENDED, OR AUTHORIZED FOR USE IN ANY MEDICAL, LIFE SAVING OR LIFE SUSTAINING SYSTEMS, OR FOR ANY OTHER APPLICATION IN WHICH THE FAILURE OF THE MATERIALS COULD CREATE A SITUATION WHERE PERSONAL INJURY OR DEATH MAY OCCUR."

and, in addition, see the following from the EULA:

You agree to maintain as confidential all information relating to your use of the Materials and not to disclose to any third party any benchmarks, performance results, or other information relating to the Materials comprising the pre-release.

Ultimately, if you have concerns, you should have a lawyer review the agreement.

[jeffrey-gallagher]

Hi Jon, nice to see your post.

Did you try here?

http://www.intel.com/software/products/vtune/beta/vt_pa_linux_beta_external.htm

Let me know what you see, as while the beta period is "formally" over, lots of folks are still kicking the tires on the product, and we want you to, as well.

Try and report back soonest.

cheers

jdg

[fons-rademakers]

Hi,

just got around installing my beta but was immediately hit by the non support of the 2.6 kernel (2.6.8.1 in my case). Isthere a patch?

Also what is this ntd, paulad and rpcss daemon stuff? What do we need all these daemons for?

Cheers, Fons.

[David_A_Intel1]

The beta does not, and willnot, support 2.6 kernels. Sorry. :-(

And, you don't need the daemons, but the analyzer does! :-} They are part of a DCOM layer used by the analyzer.

[jeffrey-gallagher]

Nice to hear from you jon.nall,

VTune shows managed code alongside unmanaged code: we just have to listen differently -- so to speak -- when we deal with unmanaged code in the mix.

Absolutely you'll see both, as long as you tell VTune that you're working with managed code as you set up the session. (Very easy to do, a single click on the data collector setup if you're using the GUI; if you're using the command line, you'll use one of these options: java, applet, or jitprofiling -- for more on that, do a man page on "sampling" or read the User's Guide.)

cheers

jdg

Message Edited by jdgallag on 11-01-2004 02:26 PM

[schorscherl]

Hi all,

when I run my OpenMP program using the cli version
of VTune 3.0b, it runs with one thread when I set OMP_NUM_THREADS=1,
as expected (this is for callgraph profiling). When I do the same
under GUI (vtlec) control, it insists on running with 2 threads. I
have tried setting OMP_NUM_THREADS=1 on the command line before
starting vtlec, as well as setting it in the 'Advanced' project
options (user-defined environment), but this does not help. OMP_NUM_THREADS
seems to get ignored when the GUI is running.

Any ideas?

Georg.

PS: This is on a dual Xeon, SuSE 8.1, Kernel 2.4.24.

[TimP]

Sometimes, it has been necessary to run a script, in which the environment variables are set before starting the application.

OpenMP programs built with Intel compilers have 2 monitor threads, in addition to those counted under OMP_NUM_THREADS. Intel libguide and linux libpthread each are responsible for one. They normally accumulate well under a second together. I'm not certain from your description whether those are involved.

[schorscherl]

Hi,

I can see the two helper threads, but there is one additional thread
which collects a lot of CPU time, and it is definitely a compute
thread. Besides, the program uses omp_get_num_threads() to determine
the number of threads it is running on, and this number is reported
as 2.

The script trick works, ok.

Btw, I have checked that the problem does not appear when doing
IP sampling. Strange.

Thanks,
Georg.